- 上一篇：Android应用程序英文文献和中文翻译
- 下一篇：液压缸泄漏诊断英文文献和中文翻译
Nielsen [7] applied dynamic programmingin the multi-criteria design optimization based on the follow-ing four performance criteria: thermal load, daylight availability,construction cost and useable area. Hauglustaine and Azar [8]optimized the building envelope using genetic algorithms; thatis, ten (10) criteria related to code compliance, energy con-sumption, and cost were considered. Wright et al. [9] appliedmulti-objective genetic algorithm to building thermal optimiza-tion with more emphasis on mechanical system design; Operatingenergy cost and occupant thermal comfort were the two per-formance criteria used. Similarly, in the case of maintainingindoor thermal comfort conditions, some scholars have alsoused multi-objective optimization algorithm to minimize build-ing energy consumption, and have achieved satisfactory results[10–12].From the above-mentioned review, it can be observed that theoptimal solutions can be achieved under the multi-objective con-straints, which illustrate that the multi-objective genetic algorithmis a feasible method for building design optimization.The current studies, however, pay more attention on using heat-ing or cooling system to improve the indoor thermal comfort,mainly focusing on resolving the conflict between building systemenergy consumption and indoor environment, without consideringhow to use building elements for efficient building energy con-sumption at the building conceptual design stage. This is regardedas an energy efficiency method. This study mainly considers theuse of building passive design to improve indoor thermal environ-ment conditions, and eliminate the factors that affect energy savingin buildings and the improvement of indoor thermal environment.This is a positive energy efficient method.2. Multi-objective optimization methodThis involves the application of an improved multi-objectivegenetic algorithm (NSGA-II), a simulation-based improved Backpropagation network, and the Pareto solution to obtain a build-ing design multi-objective optimization model. A case study ispresented to validate the multi-objective optimization model bymaintaining a wide range of trade-offs between thermal comfortand energy consumption.The multi-objective genetic algorithm is different from geneticalgorithm in that it comprehensively evaluates each objective valueof optimization solutions. However, these objective values are oftenconflicting and as a result it is difficult to find an optimal solution foreach objective. Therefore, the Pareto solution has been introducedinto this study in order to obtain a set of optimization solutions tosolve the problem.2.1. Multi-objective evolutionary algorithm (NSGA-II) as theoptimization engineThe genetic algorithm (GA) was developed by Holland in the1970s. This optimization algorithm was inspired by Darwin’s the-ory of natural selection and Mendel’s theory of genetics. It is ahighly parallel, random and adaptive optimization algorithm whichis based on “natural selection and survival of the fittest” [13].Genetic algorithms are initiated by selecting a population of ran-domly generated solutions for the problem considered. They movefrom the generation of one solution to another by evolving newsolutions using the objective evaluation, selection, crossover andmutation operators. Generally, in the genetic algorithm the solu-tions are represented by a code rather than the initial variables.Typically, a solution is represented by a string of bits (also calledchromosome). Each bit position is called gene, and the values thateach gene represent are called alleles. Nowadays, with the develop-ment of computer technology, the GA has been extensively used inmany areas, pattern recognition, image processing, neural network,optimal control etc., [14]. The GA has also been applied in severalbuilding studies, including online optimization [15], optimizationof HVAC system controls [16] and optimization of green buildingdesign [17]. In these studies, it has been proven that GA is very effi-cient even with non-differentiable functions, and in comparisonwith the baseline situation it has shown significant improvementsin the optimization result.A specific class of GA, multi-objective evolutionary algorithm(MOEA), is based on Pareto-dominance, which enables the algo-rithm to simultaneously optimize all the objectives.